Tracheal pathfinding into the CNS - CMGM Stanford - Stanford University

1505

Development 126, 1505-1514 (1999) Printed in Great Britain © The Company of Biologists Limited 1999 DEV7708

adrift, a novel bnl-induced Drosophila gene, required for tracheal pathfinding into the CNS Camilla Englund1, Anne E. Uv1, Rafael Cantera2, Laura D. Mathies3, Mark A. Krasnow4 and Christos Samakovlis1,* 1Umeå Center for Molecular Pathogenesis, Umeå University, S-90187 Umeå, Sweden 2Department of Zoology, University of Stockholm, S-10691 Stockholm, Sweden 3Department of Biochemistry, University of Wisconsin, Madison, WI 53706, USA 4Department of Biochemistry, Stanford University School of Medicine, Stanford, CA 94305,

USA

*Author for correspondence (e-mail: [email protected])

Accepted 14 January; published on WWW 3 March 1999

SUMMARY Neurons and glial cells provide guidance cues for migrating neurons. We show here that migrating epithelial cells also contact specific neurons and glia during their pathfinding, and we describe the first gene required in the process. In wild-type Drosophila embryos, the ganglionic tracheal branch navigates a remarkably complex path along specific neural and glial substrata, switching substrata five times before reaching its ultimate target in the CNS. In adrift mutants, ganglionic branches migrate normally along the intersegmental nerve, but sporadically fail to switch to the segmental nerve and enter the CNS; they wind up meandering along the ventral epidermis instead. adrift

encodes a novel nuclear protein with an evolutionarily conserved motif. The gene is required in the trachea and is expressed in the leading cells of migrating ganglionic branches where it is induced by the branchless FGF pathway. We propose that Adrift regulates expression of tracheal genes required for pathfinding on the segmental nerve, and FGF induction of adrift expression in migrating tracheal cells promotes the switch from the intersegmental to the segmental nerve.

INTRODUCTION

a secondary branch, as do the leading cell or cells of other primary branches. These secondary branches grow toward different target tissues and later ramify into the terminal branches that supply the targets with oxygen (Samakovlis et al., 1996a; Manning and Krasnow, 1993). The critical factor that controls the outgrowth of primary branches is the product of the branchless (bnl) gene, a homologue of vertebrate FGFs (Sutherland et al., 1996). bnl is expressed in discrete clusters of ectodermal and mesodermal cells arrayed around each tracheal sac. The secreted growth factor functions as a chemoattractant that activates the Breathless (Btl) FGF receptor (Klämbt et al., 1992) on nearby tracheal cells and guides their migrations as they grow out and form primary branches. Branchless signalling also induces expression of genes, including pointed (Scholz et al., 1993) and DSRF (Guillemin et al., 1996), required for secondary and terminal branching in the leading cells of the primary branches, and these cells go on to form secondary and subsequently terminal branches. The signals that guide outgrowth of secondary and terminal branches are unknown. In this paper, we investigate the cellular and molecular mechanisms that guide outgrowth of secondary ganglionic branches that grow into the ventral nerve cord. We show that these branches navigate a remarkably complex and tortuous path, tracking along five different neural and glial substrata to

Developing neurons undergo complex migrations to construct a functional nervous system. An important area of neurobiological research is aimed at identifying and characterizing the areas that guide these migrations. Many such cues are presented by developing neurons and glial cells, and the nervous system provides a highly diverse array of positional cues. This positional information potentially could be used by other migrating cells. In this paper, we demonstrate that migrating tracheal cells in Drosophila contact specific neural and glial cells as they navigate to precise targets in the CNS, and we describe a novel tracheal gene called adrift that is required for the process. The Drosophila tracheal system is a branched network of epithelial tubes that transports oxygen throughout the body. It develops from 20 ectodermal cell clusters, each containing approximately 80 cells. Each cluster invaginates, forming an epithelia sac, and then primary, secondary and terminal branches sequentially bud from each sac to generate the 20 tracheal hemisegments (Fig. 1A). Initially, six multicellular buds form and grow out along stereotyped paths to generate the six primary branches. One of these, the ganglionic branch (GB), which is the focus of this paper, targets the ventral nerve cord. The leading cell of the GB forms a unicellular tube called

Key words: Cell migration, FGF, Epithelial-CNS interaction, adrift, Drosophila, branchless

1506 C. Englund and others reach their ultimate CNS target. We identify mutations in a gene, which we call adrift, that affect the first substratum switch causing failure of GBs to enter the CNS and misrouting along the ventral epidermis. Molecular characterization of the adrift gene demonstrates that it encodes a novel nuclear protein. Expression of the gene is induced in the lead cell of the migrating ganglionic branch by the bnl FGF pathway, whereupon it promotes the switch from one neural substratum to another.

MATERIALS AND METHODS Drosophila strains The Pantip-4 P[lacZ] enhancer trap line (5405) was generated in the laboratory of M. Scott. Excision alleles of this P transposon were generated as described (Robertson et al., 1988). The 170 excision alleles obtained were balanced over CyOftzlacZ and then screened for lethality and for embryonic tracheal phenotype after staining with mAb2A12 antibody. The screen yielded several homozygous viable revertants and the two adrift alleles described here, aft1 (excision 70) and aft2 (excision 28). Other P[lacZ] enhancer trap markers used were trachealess-lacZ (1-eve-1; Perrimon et al., 1991), breathless-lacZ (T1; Bier et al., 1989), pointed-lacZ (pnt7825/Pantip1; Samakovlis et al., 1996a), sprouty-lacZ (spry9143/Pantip-2; Samakovlis et al., 1996a) and repo-lacZ (repo2138; Xiong et al., 1994). The following null or strong loss-of-function alleles were used: bnlP1 (Sutherland et al., 1996), btlLG19 (Klämbt et al., 1992), pnt∆88 (Scholz et al., 1993) and prunedex84 (Guillemin et al., 1996). Strains used for misexpression studies using the GAL4/UAS system (Brand and Perrimon, 1993) were: UAS-bnl B4-2 (Sutherland et al., 1996), btl-GAL4 (Shiga et al., 1996) and hsGAL4 (described on Flybase). Immunostaining and in situ hybridization of embryos Embryo fixation, staining and light microscopy were as described (Samakovlis et al., 1996b). The tracheal lumen-specific antibodies used were rabbit polyclonal antiserum TL-1 (from J. O’Donnell) diluted 1:2000 and mAb2A12 diluted 1:5. The anti-DSRF monoclonal antibody was mAb2-161 (1:1000) from M. Gilman (Ariad Corporation, Boston, MA). Rabbit polyclonal antibody against βgalactosidase (Cappel) was used at 1:1500. mAb22C10 against peripheral nervous system (Fujita et al., 1982; Grenningloh et al., 1991) was used 1:50. Biotin-, and Cy3-conjugated secondary antibodies (Jackson Laboratories) were used at 1:300 dilution. To prepare the Adrift antiserum, a DNA fragment from aft cDNA 3 that encodes amino acids 371-700 was inserted in frame to the 6xHis coding sequence in pRSETB (Invitrogen). The histidine-tagged Adrift fusion protein was expressed in E. coli, purified on Ni-NTA resin (QIAGEN) and used for immunization of rabbits at Agrisera (Umeå). The antiserum was affinity purified on Ni-NTA resin coupled with the histidine-tagged Aft fusion protein and used at 1:5000 final dilution for western blots and immunohistochemistry. In situ hybridization of whole-mount embryos (Lehmann and Tautz, 1994) was done with random-primed, digoxigenin-labeled aft cDNA as probe. Similar results were obtained with the 1.2 kb EcoRI aft genomic fragment. In situ hybridisation followed by antibody staining was as described (Kopczynski et al., 1996). Embryo staging was according to Campos-Ortega and Hartenstein (1985). To examine the effects on adrift expression of ectopic branchless expression, UAS-bnlB4-2;Pantip-4/hsGAL4 embryos were collected for approximately 4 hours at 21°C, aged for 7 hours at 21°C, heatshocked once for 20 minutes, and then aged 9 hours at 21°C. After fixation, embryos were stained with mAb2A12 and anti-βgalactosidase antibodies.

Electron microscopy Late stage 16 embryos were prepared for EM as described (Tepass and Hartenstein, 1994) except that the fixative was 25% glutaraldehyde, 4% paraformaldehyde and 3% tannic acid in 0.1 M phosphate buffer. Specimens were viewed by a JEOL 100 CX electron microscope. Serial sections of three different abdominal tracheal metameres from two embryos were examined. Molecular biology Genomic DNA flanking the Pantip-4 P[lacZ] insert was obtained by plasmid rescue in E.coli after cleavage of the genomic DNA with EcoRI or BamHI. Northern blots of poly(A)+ RNA, screening of genomic and cDNA libraries and sequencing were done according to Sambrook et al. (1989). Restriction mapping of 6 positive clones from a cosmid library (gift from J. Tamkun) resulted in a genomic map of 60 kb around the insertion site of the P-element. The aft cDNA 2A, a 1597 nt partial cDNA clone, was isolated from a lambda ZAP embryonic cDNA library (from P. Hurban) using the 1.2 kb EcoRI aft genomic DNA fragment as probe. Subsequently, cDNA 2A was used to probe a pupal library (from S. Stowers) and identified aft cDNA 3, a 2248 nt cDNA clone. ExoIII deletions of this cDNA were generated and used as templates for dideoxy nucleotide sequencing. cDNA 3 was used to screen an embryonic plasmid cDNA library (Brown and Kafatos, 1988), which yielded a slightly longer aft cDNA 10 (2324 nt). This cDNA was sequenced on both strands using primers based on the cDNA 3 sequence. To define intron/exon boundaries, the relevant genomic fragments were subcloned into pBluescript and sequenced. Approximately 106 phage or colonies were screened from each of the libraries. Molecular analysis of the aft mutations Genomic DNA was prepared from aft1 and aft2 homozygous flies, the parental strain and wild-type (Canton-S) flies and characterised on genomic Southern blots hybridized with fragments from cosmids surrounding the P-element insertion site and covering the aft gene. A higher resolution analysis was conducted by PCR of this genomic DNA using primers specific for the region flanking the P-element insertion site (primer A: 5′-CCTTCGACGGCGATGATCGTC-3′ and primer B: 5′-GCGGAGAACGTAACTCAAGAC-3′ deriving from wild-type cosmid sequences located 405 nt and 933 nt, respectively, upstream from the first nucleotide found in the aft cDNA). The PCR products were purified and directly sequenced using the same primers. Sequence comparisons between PCR amplified fragments from the aft1 mutant and from the corresponding region in wild-type revealed that in aft1 there is a 23 bp insertion deriving from the P-element in addition to a duplication of 8 bp of genomic sequence. In aft2 there is a 1.2 kb insert remnant from the P-element. Transgenes and misexpression studies To construct the UAS-aft transgene, a NotI-KpnI fragment from pBluescript-aft cDNA 3 plasmid was cloned into the pUAST vector (Brand and Perrimon, 1993). For the hs-aft transgenic fly strain, a NotI-XhoI fragment from the same plasmid was cloned into pPCaSpeR-hs vector (from C. Thummel). Strains carrying the transgenes were obtained by P-element mediated transformation of w1118 strain (Spradling, 1986). Two independent strains, with single insertions on the second and third chromosome respectively, were characterized for each of the two constructs. For the rescue experiment, a btl-GAL4 driver strain (Shiga et al., 1996) was used that expresses GAL4 in all tracheal cells from stage 11. Embryos of the cross aft1/CyOhblacZ;UAS-aft/+ × aft1/CyOhblacZ;btlGAL4/+ were collected at 21°C for 8 hours then transferred to 29°C for 8 hours to maximize GAL4 activity. Embryos were then fixed and stained for Adrift, tracheal lumen and βgalactosidase to score affected GBs in homozygous mutant embryos that express ectopic Adrift. To analyze the Adrift protein on western blots, wild-type embryos

Tracheal pathfinding into the CNS 1507 and embryos carrying the hs-aft construct were collected for 16 hours at 21°C, heatshocked for 45 minutes at 37°C and allowed to recover for 1 hour at 18°C. Embryos were dechorionated and homogenized in sample buffer. Electrophoresis and immunoblotting were performed according to Sambrook et al. (1989). The anti-Aft antibody was diluted 1:5000 for immunoblots and hybridizing bands were detected with the Amersham ECL reagent according to the manufacturer.

RESULTS Pathfinding of the ganglionic tracheal branch in the CNS Cell-specific tracheal and CNS markers were used to define the movements and contacts of the migrating tracheal ganglionic branches (GB) as they grow into the ventral nerve cord (VNC). There are twenty ganglionic branches, two in each Tr1 tracheal hemisgement and one in each hemisegment from Tr2 to Tr9 (Fig. 1B) (Manning and Krasnow, 1993). The navigation pathway of each GB is complex but highly stereotyped, involving interactions with two different nerves and several glial cell types along the way. The path followed by the 14 abdominal GBs is the same in each segment; it is summarized in Fig. 1J and described in detail below. The path followed by the six thoracic GBs is similar except they turn dorsally and anteriorly as they approach the midline in the VNC (not shown). The typical GB is an epithelial tube composed of seven cells (Samakovlis et al., 1996a). As the GB buds and grows out, a single cell called GB1 at the end of the branch leads the migrations (Fig. 1A,C); the remaining six GB cells (GB2-GB7) follow in line behind (Samakovlis et al., 1996a). At embryonic stage 13, the GB cells migrate ventrally towards the CNS, in close contact with the intersegmental nerve (ISN) (Fig. 1D and data not shown). The GB cells are attracted towards the CNS by two ventrolateral clusters of epidermal cells expressing the bnl FGF (Sutherland et al., 1996). At stage 14, as the tip of the GB approaches the entry point into the VNC, expression of bnl ceases. The lead cell (GB1) turns posteriorly, breaks contact with the ISN and changes substratum to associate with the segmental nerve (SN) (Fig. 1D). During stage 15, GB1 migrates along the SN as it traverses laterally and ventrally inside the VNC, GB2 follows GB1 onto the SN. The other GB cells (GB3-GB7) also release contact with the ISN but they do not bind the SN or enter the VNC and they end up suspended in the body cavity (Fig. 1C). A group of CNS glial cells called exit glia (see Goodman and Doe, 1993 for nomenclature) surround the point where the GB cells switch from the ISN to the SN, suggesting that they may play a role in the switching (Fig. 1E). As the GB1 cell continues medially on the ventral side of the neuropil during stage 16, it comes in close contact with the ventral longitudinal glial cells (Fig. 1F). Just before reaching the midline glia, GB1 turns abruptly and migrates dorsally (Fig. 1G) to reach the dorsal side of the neuropil. It then turns again and extends posteriorly on the dorsal longitudinal glial tracts toward the neighboring hemisegment (Fig. 1G; data not shown). EM analysis of sagittal sections of the VNC of stage 16 wildtype embryos confirmed the direct association between GB cells and various glial cells along the migration pathway. Inside

the VNC, just beyond the point of the ISN to SN switch, a segmental nerve root glial cell (SNG) (Ito et al., 1995) completely engulfs both the GB and the SN (Fig. 1H). In serial sections that allow one to follow the GB towards its tip, contacts between GB1 and longitudinal glia and dorsoventral channel glia are also seen (Fig. 1I; data not shown). Ganglionic branches are misrouted in adrift mutants The Pantip-4 P(lacZ, w+) transposon is located at chromosomal position 54F and expresses the β-galactosidase marker in the lead cells of the GB and other growing primary branches (Samakovlis et al., 1996a). The original insertion caused weak, sporadic defects in GB outgrowth (not shown). By introducing a source of transposase, 170 w− transposon excision alleles were obtained. Two homozygous viable alleles (excisions 28 and 70) displayed a similar tracheal phenotype and failed to complement for this function. We refer to the alleles as adrift1 and adrift2 because of their tracheal phenotypes described below. Molecular analysis indicates that aft1 represents the zygotic null condition for the tracheal function of the gene (see below) and was the focus of the phenotypic analysis. In aft mutants, GBs sporadically stalled or were misrouted and failed to follow their stereotyped path into the CNS. In aft1 mutants, 18% of GBs (n=300) missed the entry point into the VNC and continued to migrate along the ventral epidermis (Fig. 2A; see also below). An additional 6% of GBs stalled at or before reaching the VNC. Neither defect was seen in wildtype controls (n=400). Double labeling with tracheal and neural markers demonstrated that GBs in homozygous aft1 embryos grew normally over the ISN during the initial phase of their ventral migration but sporadically failed to make the switch to the SN and associate with the exit glial cells (Fig. 2B,C). Most of the affected branches instead turned posteriorly and continued to migrate along the ventral epidermis, forming a characteristic ‘hook’ structure (Fig. 2A, top arrow). The defect in GB guidance in aft1 mutants does not result from grossly aberrant differentiation of the tracheal cells. The cells retained their ability to migrate and expressed all of the appropriate secondary and terminal branch markers tested including pointed7825 lacZ, sprouty9143 lacZ, Pantip-4 lacZ and DSRF (Fig. 2D,E; data not shown). Furthermore, the structure of the nerves and glial cells normally contacted by the growing GB were unaffected in the mutant (Fig. 2B,C; data not shown). We conclude that aft is specifically required to promote recognition and association of the GB1 cell with the SN and glial cells at the first guidepost in its navigation into the CNS.

adrift encodes a nuclear protein with a novel, evolutionarily conserved motif Genomic DNA fragments flanking the Pantip-4 P-element insertion were recovered by plasmid rescue in E. coli and used to probe a wild-type Drosophila genomic library. Seven overlapping cosmids were obtained and a genomic map covering 60 kb surrounding the transposon insertion site was constructed (Fig. 3A). Genomic fragments near the insertion site were used to probe northern blots of embryonic mRNA. A 2.3 kb transcript located ~0.7 kb from the insertion site was identified (Fig. 3A,B) and found to be expressed in the trachea in the same pattern as the lacZ reporter in the Pantip-4 strain (see below). This transcription unit corresponds to the aft gene because its expression was reduced or absent in the aft mutants

1508 C. Englund and others Fig. 1. Pathfinding of the tracheal ganglionic branch into the CNS. (A) Diagram of two abdominal tracheal hemisegments at embryonic stage 15 (lateral view, anterior left). The primary branches are labeled: DB, dorsal branch; DT, dorsal trunk; VB, visceral branch; LT, lateral trunk; GB, ganglionic branch. The nucleus of the GB terminal cell and other terminal cells that lead the outgrowth of primary branches and later form secondary and terminal branches are indicated by the black circles. The remaining 6 cells of the GB are indicated by the open circles. The cells are named (GB1GB7) according to their position in the branch starting with the terminal cell (GB1, black circle). (B) Ventral view of a wild-type stage 16 embryo stained for mAb2A12 antigen in the tracheal lumen (brown). The plane of focus is on the ventral nerve cord to show the 20 GBs that have migrated into the CNS. The six thoracic GBs are marked by asterisks. (C-G) Stage 15 or 16 embryos carrying one copy of the indicated lacZ enhancer trap markers were double stained for a tracheal and a neural or glial marker. Close up of the tips of one (C) or two GBs (marked with concave arrowheads, D-G) are shown as they migrate into and through the CNS (ventral view, anterior left). (C) Stage 16 embryo carrying a btllacZ reporter (T1) double stained for βgal to show nuclei of tracheal and CNS midline cells (both brown) and mAb2A12 antigen to show tracheal lumen (also brown). The nucleus of the GB1 cell is in the CNS and the GB2 nucleus lies just outside. The CNS border is outlined by black dots. (D) Stage 15 embryo carrying a trhlacZ reporter (1-eve-1) double stained for β-gal to show tracheal cells (brown) and mAb22C10 antigen to show nerves (blue). GBs veer posteriorly just before entering the CNS, releasing contact with the intersegmental nerve (ISN) and associating with the segmental nerve (SN). (E,F) Embryos carrying a repo-lacZ reporter (repo2138) double stained for β-gal to show glial cells (brown) and mAb2A12 antigen to show tracheal lumen (blue). On their way into the CNS at stage 15 (E), each GB is in close contact with exit glial cells (EG). Two hours later at stage 16 (F), each GB migrates across the ventral side of the longitudinal glia (LG). (G) Stage 16 embryo stained as in C. When the GB reaches the midline glial cells (MG), it turns and migrates dorsally through the dorsoventral channel. (H,I) Electron micrographs of sagittal sections of stage 16 CNS showing contacts between GB cells and CNS cells. Cell contacts are outlined (broken lines). (H) At the entry point into the CNS the GB2 cell, the segmental nerve root (SN) and the intersegmental nerve root (ISN) are all enveloped by a segmental nerve root glial cell (SNG). (I) As the GB migrates dorsally along the dorsoventral channel, the GB1 cell is in direct contact with channel glial cells (ChG). (J) A simplified transverse projection of the nerve cord (grey) in a stage 16 embryo to illustrate the position of the GB (orange/brown) in the CNS. The steps in GB migration are as follows. (I) The GB switches substratum from the ISN to the SN root just before entering the ventral nerve cord. (II) GB migrates into the CNS along the SN. GB2 (orange) and the SN are both enveloped by a SN root glial cell. (III) GB1 (brown) extends towards the ventral surface of the neuropil (NP) to associate with the ventral longitudinal glial cells; GB2 remain at the CNS periphery. (IV) At the ventral midline, GB turns and migrates dorsally between the two neuropils in close contact with channel glial cells. (V) When GB reaches the dorsal side of the neuropil, it associates with the dorsal longitudinal glial cells and extends laterally and posteriorly. Bar in B, 50 µm; bar in C-G, 10 µm; bar in H and I, 0.7 µm.

Tracheal pathfinding into the CNS 1509

Fig. 2. Ganglionic branches are misrouted in aft mutants. (A) Stage 16 aft1 mutant embryo stained with mAb2A12 to visualize tracheal lumen. GBs sporadically fail to enter the CNS and instead migrate posteriorly forming ‘hooks’ (arrows) or stall near the entry site (arrowheads). The hooks on the bottom curve are out of the focal plane. The most posterior pair of GBs are outside the focal plane. (B) Close-up of two GBs in a stage 15 aft1 mutant carrying one copy of a trh-lacZ reporter (1-eve-1). The embryo was stained for β-gal to show tracheal cells (brown) and mAb22C10 antigen to show nerves (blue). The misrouted GB (arrowhead) released the ISN as in wild type (see Fig. 1D) but failed to attach to the SN (arrow). (C) Close up of two GBs in a stage 15 aft1 mutant carrying one copy of a repolacZ reporter (repo2138). The embryo was stained for β-gal to show glial cells (brown) and mAb2A12 antigen to show tracheal lumen (blue). The misrouted branch turns posteriorly at a point close to the exit glial cells (just below focal plane at arrowhead) but is not associated with these cells as are wild-type GBs (see Fig. 1E). (D) Stage 16 aft1 mutant stained with mAb2A12 (blue) and mAb against the tracheal terminal cell marker DSRF (brown). The GB1 cell (arrowhead) in this misrouted branch expresses the marker normally. (E) Stage 16 aft1 mutant carrying one copy of a spry-lacZ reporter (spry9143) and stained for β-gal (brown) and mAb2A12 (blue). The terminal cell (arrowhead) expresses the spry marker normally. All panels show ventral or ventrolateral views (anterior left). Bar in A, 50 µm; bar in B-E, 10 µm.

and the tracheal phenotype of aft mutants could be rescued by expression of a representative cDNA (see below). Three cDNA clones corresponding to the identified transcription unit were obtained by screening embryonic and pupal cDNA libraries. Restriction mapping and sequence analysis of the cDNAs and the corresponding genomic region established the intron-exon structure shown in Fig. 3A. The longest cDNA contains a 700 residue long ORF (Fig. 3C) and an NCBI BLAST search of the combined databases identified several related cDNAs and Expressed Sequence Tags from different species. All show 18-28% identity to a 184 residue region in the N-terminal half of the predicted Adrift protein (residues 135-319; Fig. 3D). We call this the Adrift domain. The only other characterized gene containing an Adrift domain is the yeast YCL54 gene. Temperature-sensitive mutants in YCL54 were isolated in a screen for mutants defective in

mating type silencing in yeast (Loo et al., 1995) but the molecular function of YCL54 protein in the process is unknown. There were no other significant similarities between the predicted Adrift ORF and the sequences deposited in the databases. To examine the subcellular localization of Adrift protein, we raised polyclonal antisera against a recombinant form of the protein consisting of the C-terminal half of Adrift fused to a six histidine epitope tag. On western blots of Drosophila embryo extracts, the affinity-purified antiserum recognized a single 72 kDa species, in good agreement with the calculated molecular weight. This species was strongly induced in extracts of heat-shocked embryos carrying a construct including the aft cDNA3 under the control of the hsp70 promoter, confirming that it is the Adrift protein (data not shown). Immunolocalization studies in wild-type embryos detected Aft protein in the nuclei of all cells at cellular blastoderm stage (Fig. 4A,B), presumably derived from the abundant maternal aft mRNA (see below), and also in nuclei of the gonads, epidermis and brain lobes of older embryos (Fig. 4C,D). Unfortunately, the antisera were not sensitive enough to reproducibly detect the endogenous tracheal expression of the gene. However, when aft was overexpressed in the developing tracheal system using the UAS/GAL4 system, Aft antigen was readily detected and was predominantly nuclear (Fig. 4E).

adrift is expressed dynamically and required in migrating ganglionic branches The pattern of adrift transcription during embryonic development was determined by northern blot analysis, wholemount in situ hybridization and analysis of an adrift lacZ reporter (Pantip-4). The gene is maternally expressed and transcripts are evenly distributed during the syncytial blastoderm stage; most of the maternal transcripts are degraded during early embryogenesis (Fig. 3B and data not shown). Zygotic transcription is first detected by in situ hybridization in midembryogenesis in the developing tracheal system and later in the developing gonad (Fig. 5B,J). Tracheal expression is highly dynamic as revealed by both in situ hybridization and expression of the adrift lacZ reporter. At stage 11, the gene is expressed weakly in all tracheal cells (not shown). As the primary branches bud and grow out during stages 12 and 13, the gene is preferentially expressed in the leading cells of the GB and other growing primary branches (Fig. 5A,B). During stages 14 and 15, expression becomes further restricted to just the GB1 terminal cell (Fig. 5C,I) and other terminal cells (Fig. 5E-H) that lead the migrations toward the CNS and other target tissues. No tracheal expression was detected in aft1 mutant embryos (Fig. 5D) and expression was severely reduced in aft2 mutants (data not shown). Maternal and gonadal expression, however, were detected in the mutants (Fig. 5K and data not shown), indicating that aft1 and aft2 selectively affect tracheal expression of the gene. Analysis of the molecular lesions in the alleles identified alterations at the P-element insertion site upstream of the coding region, suggesting that this may be a cis-regulatory region important for tracheal expression of the gene (see legend to Fig. 3). An aft transgene expressed specifically in the developing tracheal system was constructed and used to confirm that aft function is specifically required in the migrating tracheal cells. A

1510 C. Englund and others UAS-aft transgene was crossed to a btl-GAL4 driver line to drive aft expression in all tracheal cells from stage 11 onwards (Shiga et al., 1996). Tracheal expression of the aft transgene almost completely rescued the misrouting phenotype of homozygous aft1 mutant embryos. 24% of GBs are affected in aft1 mutants whereas only 1% of GBs were affected in aft1 mutants expressing the adrift transgene (Table 1). Thus, aft is required in migrating tracheal cells to find their targets in the CNS.

adrift expression is regulated by the Branchless FGF pathway adrift is expressed in the leading cells of growing tracheal branches, near clusters of branchless FGF-expressing cells and in a pattern very similar to that of several known branchless-induced genes including pointed (Samakovlis et al., 1996a), DSRF/pruned (Samakovlis et al., 1996a) and sprouty (Hacohen et al., 1998). This suggested that adrift expression might also be induced by the bnl signalling pathway. We examined expression of the adrift lacZ reporter in embryos mutant for four components of the branchless pathway: bnl (Sutherland et al., 1996), btl (Klämbt et al., 1992), pnt (Scholz et al., 1993) and pruned (Guillemin et al., 1996). Initial expression of the adrift reporter in stage 11 tracheal cells was normal in all four mutants (not shown), but subsequent expression in the leading cells of the branches was absent in bnl, btl and pnt mutants (Fig. 6B-D). Expression in pruned mutants was unaffected (data not shown). In a complementary experiment in which bnl was misexpressed under the control of the hsp70 promoter, expression of the adrift reporter expanded to include additional cells in each branch (Fig. 6E). Thus, the Branchless FGF pathway induces aft expression in the leading cells of tracheal branches, and this induction requires the bnl FGF, the btl FGF receptor and the pointed ETS domain transcription factor. DISCUSSION GB tracheal cells navigate a complex pathway to reach their targets in the CNS We have shown that the GB1 cell at the end of each migrating GB reaches the CNS at stage 15 and migrates more than 50 µm over the next 5 hours, following a remarkably complex path comparable to the most elaborate paths known in axonal pathfinding in Drosophila. We can conceptually divide the GB migration path into five segments based on the direction of migration and the migration substratum and cell contacts (see Fig. 1J). Initially, the GB tracks along the ISN towards the CNS (segment I). At the CNS entry point, the GB switches onto the SN and associates with exit glia nearby. In the CNS, the GB1 cell grows along the SN and segmental nerve root glia (segment II) and then the ventral longitudinal glia (segment III), approaching the midline. At the midline, it turns and extends dorsally along the dorsoventral channel surrounded by the channel glia (segment IV). Once it reaches the dorsal side of the neuropil, the GB1 cell turns back away from the midline, contacting the dorsal longitudinal glia and then continuing migration posteriorly along the longitudinal connectives and surrounding glia towards the next hemisegment (segment V), completing its embryonic migrations. During larval life, GB1 grows again, this time ramifying extensively into fine terminal

branches that provide the ~230 neurons in each hemisegment with oxygen (E. Johnson, J. Jarecki, and M. A. K., unpublished data). Each segment of this journey is presumably governed by distinct signalling events between the tracheal cell and its immediate environment. The only signal identified so far is the branchless FGF, which is expressed in discrete clusters of epidermal cells and attracts the GB toward the CNS during the first segment of the journey. The signals that guide the subsequent migrations into and around the CNS are unknown. Once in the CNS the GB1 cell comes in close contact with various neurons and glia. It thus is exposed to many of the same signals that guide axonal outgrowth and some of these could be used to guide GB migrations in the CNS. Indeed, several cell surface molecules implicated axon guidance, including Fasciclin II and Neuroglian, are also expressed on the surface of growing GB cells (Bieber et al., 1989; Van Vactor et al., 1993). Fig. 3. Structure of the aft locus and gene product. (A) Map of the aft locus. The intron-exon structure of the aft cDNA and a neighboring, overlapping gene (ubiquitin-conjugating enzyme, ubc) are shown below the genomic map; the shaded areas are coding regions. The position of the P[lacZ] insert in Pantip-4 is indicated (large triangle); the arrow shows the direction of lacZ transcription. Sequencing of genomic DNA encompassing the transposon in the Pantip-4 strain mapped the insertion site 723 bp upstream of the first nucleotide found in the sequence of the cDNA clone. Molecular defects in the excision mutants were analyzed by genomic DNA blots, using cosmid fragments deriving from the aft gene as probes, and by direct sequencing of PCR fragments deriving from genomic DNA of aft1 and aft2 flies. Sequence comparisons between PCR amplified fragments from the genomic region around the P-element insertion site in aft1 and from the corresponding region in wild-type revealed that in aft1 there is a 23 bp insertion deriving from the P element in addition to a duplication of 8 bp of genomic sequence. In aft2 there is a 1.2 kb insert remnant from the P element. The black box shows the 1.2 kb genomic EcoRI fragment used to identify the aft cDNA. R, EcoRI. (B) Northern blot of embryonic mRNA from embryos of the indicated ages probed with a radiolabeled 1.2 kb aft genomic EcoRI fragment (upper panel, 48 hour exposure) and an actin5C probe (Fyrberg et al., 1980) (lower panel, 15 hour exposure) as a loading control. The aft probe detects a single 2.3 kb transcript. Zygotic transcript is first detected at 8 and 12 hours; the high level of transcript in 0- to 4-hour-old embryos is apparently of maternal origin. (C) Sequence of aft cDNA and the predicted protein (GenBank accession number AF117649). In-frame stop codons following the coding sequence are underlined. The Adrift domain shown in D is highlighted in gray. Arrowheads indicate the intron positions. (D) Sequence similarity between 184 amino acids of the predicted D. melanogaster Adrift protein (D.m.) and portions of five other known or predicted proteins. Shaded boxes mark amino acid identities between Adrift and the others identified by FASTA3 searches (Pearson and Lipman, 1988) of the non-redundant amino acid database. Brackets mark sequence blocks conserved between all sequences as identified by a multiple sequence alignment tool (Depiereux et al., 1997). No obvious homology was detected in sequences outside this domain. A.s, African swine fever virus pEP424R (accession number U18466, 28% identity), H.s, a novel human gene (D43949, 25%), E.p, Epiphyas postvittana nucleopolyhedrovirus (AF037358, 25%), C. e, Caenorhabditis elegans R74.7 (Z36238, 20%), S.c, Saccharomyces cerevisiae hypothetical 34.7 kDa protein in ORC-TIP1 intergenic region (Z35930, 18%).

Tracheal pathfinding into the CNS 1511 The adrift gene is required in the trachea for efficient entry into the CNS One of the most challenging transitions on the GB migration path occurs at the entry into the CNS. At late stage 14, all seven

cells of the branch have migrated ventrally in response to bnl signalling, and the lead cell, GB1, has been induced to express secondary and terminal branching genes. The GB1 cell leaves the ISN and associates with the SN. This substratum switch

A 1 kb

R

R

R

R

+29.6 kb

+21.5 kb 5'

3' af t 3'

5' ub c

1512 C. Englund and others Table 1. Rescue of aft1 mutant by tracheal expression of a UAS-aft transgene Genotype

Affected ganglionic branches* (%)

OregonR (wild type) aft1/aft1 aft1/aft1;UAS-aft/btl-GAL4

0 (n=400) 24 (n=300) 1 (n=300)

*Misrouted or stalled ganglionic branches.

Fig. 4. Immunostaining of Adrift protein in wild-type embryos (A-D) and transgenic embryos overexpressing Aft (E; UAS-aft/btl-GAL4). (A) Stage 5 embryo. Aft protein is detected throughout the blastoderm. (B) Close-up of embryo in A. Aft protein is localized to the nucleus. (C,D) Close up (C) and lower magnification view (D) of stage 15 embryo (ventral view) showing Aft expression in nuclei of the gonads (arrows). In D, Aft protein is also seen in the brain lobes (arrowheads) and epidermis. (E) Stage 16 embryo (ventral view) overexpressing Aft protein in all tracheal cells and in midline cells of the CNS. Concave arrowheads indicate nuclear expression in the terminal GB1 cells of ganglionic branches. Bar in A, D and E, 50 µm; bar in B and C, 10 µm.

occurs at the CNS exit junction, which is composed of the two major nerve roots surrounded by a specialized group of exit glial cells. The junction is also a plexus region where motorneurons leaving the CNS switch pathways and selectively fasciculate into five major nerve branches destined for different regions of the musculature (Goodman and Doe, 1993; Van Vactor et al., 1993; Landgraf et al., 1997). Thus, the GB1 cell must not only release its old substratum but also select a new substratum from among several glial cells and neuronal projections. The GB2 cell makes a similar choice when it crosses the same junction ~1 hour later.

In aft mutants, most of the affected GBs migrate into the exit junction, release contact with the ISN, and extend posteriorly towards the SN but fail to attach to it or the surrounding glia. Instead, they continue migrating posteriorly along the ventral epidermis and muscle. This suggests that adrift functions in a signalling pathway that attracts the GB onto the SN or fixes it on it once there. The expression pattern of the gene and mutant rescue by an aft transgene expressed in migrating tracheal cells demonstrate that aft is required in the cells that receive the signal and not in the signalling cells or the SN cells to which the GB cells adhere. What might be the source of this signal? The most obvious candidates are the SN neurons or surrounding glial cells. Several axons including those of the RP1, RP3, RP4 and RP5 motoneurons switch from ISN to SN as they leave the CNS at the exit junction (Goodman and Doe, 1993). These might express short-range attractive signals or adhesion molecules and provide a bridge for the growth of the GB from the ISN to SN. Other potential sources of the signal are the three or four exit glia cells and the segmental nerve root glia that are intimately associated with the GB at the exit junction. In the absence of mutants that selectively affect these glial subtypes, we have not been able to specifically assess their roles in GB pathfinding. However, preliminary analysis of the glial cells missing (gcm) mutant (Jones et al., 1995; Vincent et al., 1996) in which many glial cell types fail to differentiate properly, has revealed several GB pathfinding phenotypes including misrouted branches that do not enter the CNS, supporting a role for glial cells in GB pathfinding events.

adrift function and regulation by the Branchless FGF pathway The localized and dynamic expression of the bnl FGF in the epidermis and other tissues guides tracheal cell migrations during primary branching including outgrowth of the GB towards the CNS. In addition to this chemoattractant function, the bnl FGF also induces expression of genes required for secondary and terminal branches and thus selects where the next generation of branches will sprout (Lee et al., 1996; Sutherland et al., 1996). Our results identify a third function of bnl signalling in tracheal development – induction of a gene involved in pathfinding. Induction of adrift expression at the end of the migrating GB promotes the switch from the ISN to the SN. Although we do not know at the molecular level how Adrift promotes this pathfinding switch, the nuclear localization of Adrift protein and its sequence similarity to a gene implicated in transcriptional silencing in yeast suggest that Adrift may regulate expression of other genes, perhaps one involved in recognition and adhesion to the SN or nearby glia. Adrift and the yeast protein may define a new class of regulatory proteins

Tracheal pathfinding into the CNS 1513 Fig. 5. Embryonic expression pattern of aft. aft expression was detected by in situ hybridization with an aft cDNA or by βgal antibody staining (blue) of embryos carrying one copy of the aft-lacZ marker Pantip-4 (A,C,E-G); some embryos were also stained with mAb2A12 (C,E-G) or TL-1 antisera (H,I) to show tracheal lumen (brown). (A) Stage 12/13 aft-lacZ embryo (lateral view, anterior left). The aft-lacZ marker is expressed in the ganglionic branch and other growing primary branches including the DBs (DB, dorsal branch; VB, visceral branch; GB, ganglionic branch). (B) Stage 12/13 embryo (lateral view). aft RNA is expressed in the leading cells of the GB and other growing primary branches including the DBs. (C) Stage 16 aft-lacZ embryo (ventral view, anterior left). Expression of the marker has restricted to the GB1 cells (arrowhead) and other tracheal terminal cells (not shown). (D) Stage 12-13 aft1 mutant (lateral view). No expression of aft RNA is detectable in the trachea. Compare to wild-type embryo in B which was treated identically. (E-G) Close-up views of growing dorsal branch in embryos carrying one copy of the aft-lacZ marker. Note the dynamic restriction of marker expression to the leading cells. DB, dorsal branch; DT, dorsal trunk. (E) Stage 12. The marker is expressed in all cells of the budding DB. (F) Stage 14. Expression becomes restricted to the leading cells of the DB. (G) Stage 15. Strong expression is seen only in the DB1 terminal cell. Weak residual β-gal can be detected in the other DB cells. (H) Close up view of growing dorsal branch in wild-type stage 14 embryo showing a similar restriction of aft RNA to the DB1 cell. (I) Closeup view of growing ganglionic branch in wild-type stage 14 embryo showing similar restriction of aft RNA to the GB1 cell. (J) aft RNA is detected in the gonads (arrows) of the stage 15/16 wild-type embryo (ventral view, anterior up). (K) Gonadal expression remains in the aft1 mutant (arrows). However, tracheal expression seen in the wild-type (arrowheads in J) is missing in the mutant. Bar in A-D, 50 µm; bar in E-I, 10 µm, bar in J and K, 50 µm.

as a similar conserved motif was found in several anonymous predicted proteins in other organisms from C. elegans to humans (Fig. 3D). aft expression is induced by Bnl signalling at the ends of all growing primary branches. However, aft mutations affect only

the GBs. The lack of effect in other branches could be due to persistence of a low level of maternal aft expression that is sufficient for function in these branches. Alternatively, aft might be fully redundant or have no function in other branches; even in GB pathfinding, it does not appear absolutely essential as many GBs reach their proper targets in the mutants. Another gene called unplugged (unp), which encodes a nuclear homeobox protein, is also variably required for GB entry into the CNS (Chiang et al., 1995). The perception and response to cues provided at the CNS entry site may thus require the combined action of aft and unp. Because unp is expressed in Fig. 6. The Branchless FGF pathway regulates aft expression. aft expression was monitored in different Branchless pathway mutants using the aft-lacZ marker Pantip-4. The stage 15 embryos are double stained for β-gal (blue) to show aft-lacZ expression and mAb2A12 antigen to show tracheal lumen (brown). (A) Wild type. aft is expressed at high levels in DB1 (arrow) and other terminal cells. DB, dorsal branch; DT, dorsal trunk. (B) branchless homozygote. In the absence of the branchless FGF, the DB and other primary branches fail to form and aft expression is not induced. (C) breathless homozygote. Absence of the breathless FGF receptor causes a similar phenotype. (D) pointed homozygote. In the absence of the pointed ETS domain transcription factor, a downstream effector in the Branchless pathway, the DB and other primary branches form but aft expression is not induced (arrow). The weak early expression of aft in all tracheal cells persists in the mutants (arrowheads). (E) UASbnlB42; hsGAL4 embryo. Additional DB cells (arrow) express aft after ubiquitous expression of branchless. Bar for A-E, 10 µm.

1514 C. Englund and others all cells of the GB and cerebral branches but not other branches, only the lead cells of branches that target the CNS would express the proper combination of regulators and only at the right time just before entering the CNS. We thank N. Brown, C. Goodman, J. Hurban, S, Stowers, J. Tamkun for cDNA libraries and flies. We also thank H. and T. Edlund, D. Hultmark, H. Ronne, S. Tuck, J. Yochem and members of our laboratories for comments on the manuscript. The drawing in Fig. 1J was prepared from an image generously provided by K. Ito and C. Rickert. This work was supported by grants from the Swedish Research Council (NFR) and a Junior Investigator Award from SSF to C. S.

REFERENCES Bieber, A. J., Snow, P. M., Hortsch, M., Patel, N. H., Jacobs, J. R., Traquina, Z. R., Schilling, J. and Goodman, C. S. (1989). Drosophila Neuroglian: a member of the immunoglobulin superfamily with extensive homology to the vertebrate neural adhesion molecule L1. Cell 59, 447-460. Bier, E., Vassein, H., Sheperd, S., Lee, K., McCall, K., Barbel, S., Ackerman, L., Carretto, R., Uemura, T., Grell, E., Jan, L. and Jan, Y. (1989). Searching for pattern and mutation in the Drosophila genome with a P-lacZ vector. Genes Dev. 3, 1273-1287. Brand, A. H. and Perrimon, N. (1993). Targeted gene expression as a means of altering cell fates and generating dominant phenotypes. Development 118, 401-415. Brown, N. H. and Kafatos, F. C. J. (1988). Functional cDNA libraries from Drosophila embryos. J. Mol. Biol. 203, 425-437. Campos-Ortega, A. J. and Hartenstein, V. (1985). The Embryonic Development of Drosophila melanogaster. New York: Springer-Verlag. Chiang, C., Young, K. E. and Beachy, P. A. (1995). Control of Drosophila tracheal branching by the novel homeodomain gene unplugged, a regulatory target for genes of the bithorax complex. Development 121, 3901-3912. Depiereux, E., Baudoux, G., Briffeuil, P., Reginster, I., Bolle, X. D., Vinals, C. and Feytmans, E. (1997). Match-Box_server: a multiple alignment tool placing emphasis on reliability. CABIOS 13, 249-256. Fujita, S. C., Zipursky, S. L., Benzer, S., Ferrus, A. and Shotwell, S. L. (1982). Monoclonal antibodies against the Drosophila nervous system. Proc. Natl. Acad. Sci. 79, 7929-7933. Fyrberg, E. A., Kindle, K. L. and Davidson, N. (1980). The actin genes of Drosophila: a dispersed multigene family. Cell 19, 365-378. Goodman, C. S. and Doe, C. Q. (1993). Embryonic development of the Drosophila central nervous system. In The Development of Drosophila melanogaster (ed. M. Bate, and A. M. Arias), pp. 1131-1206. Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press. Grenningloh, G., Rehm, E. J. and Goodman, C. S. (1991). Genetic analysis of growth cone guidance in Drosophila: Fasiclin II functions as a neuronal recognition molecule. Cell 67, 47-57. Guillemin, K., Groppe, J., Ducker, K., Treisman, R., Hafen, E., Affolter, M. and Krasnow, M. A. (1996). The pruned gene encodes the Drosophila serum response factor and regulates cytoplasmic outgrowth during terminal branching of the tracheal system. Development 122, 1353-1362. Hacohen, N., Kramer, S., Sutherland, D., Hiromi, Y. and Krasnow, M. (1998). sprouty encodes a novel antagonist of FGF signaling that patterns apical branching of the Drosophila airways. Cell 92, 253-263. Ito, K., Urban, J. and Technau, G. M. (1995). Distribution, classification, and development of Drosophila glial cells in the late embryonic and early larval ventral nerve cord. Roux Arch. Dev. Biol. 204, 284-307. Jones, B. W., Fetter, R. D., Tear, G. and Goodman, C. S. (1995). glial cells

missing: a genetic switch that controls glial versus neuronal fate. Cell 82, 1013-1023. Klämbt, C., Glazer, L. and Shilo, B. -Z. (1992). breathless, a Drosophila FGF receptor homolog, is essential for migration of tracheal and specific midline glial cells. Genes Dev. 6, 1668-1678. Kopczynski, C. C., Davis, G. W. and Goodman, C. S. (1996). A neural tetraspanin, encoded by late bloomer, that facilitates synapse formation. Science 271, 1867-1870. Landgraf, M., Bossing, T., Technau, G. M. and Bate, M. (1997). The origin, location, and projections of the embryonic abdominal motorneurons of Drosophila. J. Neurosci 17, 9642-9655. Lee, T., Hacohen, N., Krasnow, M. and Montell, D. J. (1996). Regulated Breathless receptor tyrosine kinase activity required to pattern cell migration and branching in the Drosophila tracheal system. Genes Dev. 10, 2912-21. Lehmann, R. and Tautz, D. (1994). In Situ Hybridization to RNA. In Methods in Cell Biology (ed. L. S. B. Goldstein, and E. A. Fyrberg), pp. 575-598. San Diego: Academic press, inc. Loo, S., Laurenson, P., Foss, M., Dillin, A. and Rine, J. (1995). Roles of ABF1, NPL3, and YCL54 in silencing in Saccharomyces cerevisiae. Genetics 141, 889-902. Manning, G. and Krasnow, M. A. (1993). Development of the Drosophila tracheal system. In The Development of Drosophila melanogaster (ed. A. M. Arias, and M. Bate), pp. 609-685. Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press. Pearson, W. R. and Lipman, D. J. (1988). Improved tools for biological sequence comparisons. Proc. Natl Acad. Sci. USA 85, 2444-2448. Perrimon, N., Noll, E., McCall, K. and Brand, A. (1991). Generating lineage-specific markers to study Drosophila development. Dev. Genet. 12, 238-252. Robertson, H. M., Preston, C. R., Phillis, R. W., Johnson-Schlitz, D. M., Benz, W. K. and Engels, W. R. (1988). A stable genomic source of P element transposase in Drosophila melanogaster. Genetics 118, 461-470. Samakovlis, C., Hacohen, N., Manning, G., Sutherland, D., Guillemin, K. and Krasnow, M. A. (1996a). Development of the Drosophila tracheal system occurs by a series of morphologically distinct but genetically coupled branching events. Development 122, 1395-1407. Samakovlis, C., Manning, G., Steneberg, P., Hacohen, N., Cantera, R. and Krasnow, M. A. (1996b). Genetic control of epithelial tube fusion during Drosophila tracheal development. Development 122, 3531-3536. Sambrook, J., Fritsch, E. F. and Maniatis, T. (1989). Molecular Cloning: A Laboratory Manual. Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press. Scholz, H., Deatrick, J., Klaes, A. and Klämbt, C. (1993). Genetic dissection of pointed, a Drosophila gene encoding two ETS-related proteins. Genetics 135, 455-468. Shiga, Y., Tanaka-Matakatsu, M. and Hayashi, S. (1996). A nuclear GFP/βgalactosidase fusion protein as a marker for morphogenesis in living Drosophila. Develop. Growth Differ. 38, 99-106. Spradling, A. C. (1986). P element-mediated transformation. Oxford: IRL Press Limited. Sutherland, D., Samakovlis, C. and Krasnow, M. A. (1996). branchless encodes a Drosophila FGF homolog that controls tracheal cell migration and the pattern of branching. Cell 87, 1091-1101. Tepass, U. and Hartenstein, V. (1994). The development of cellular junctions in the Drosophila embryo. Dev. Biol. 161, 563-596. Van Vactor, D., Sink, H., Fambrough, D., Tsoo, R. and Goodman, C. S. (1993). Genes that control neuromuscular specificity in Drosophila. Cell 73, 1137-1153. Vincent, S., Vonesch, J. -L. and Giangrande, A. (1996). glide directs glial fate commitment and cell fate switch between neurones and glia. Development 122, 131-138. Xiong, W. C., Okano, H., Patel, N. H., Blendy, J. A. and Montell, C. (1994). repo encodes a glial-specific homeo domain protein required in the Drosophila nervous system. Genes Dev. 8, 981-994.